Title :
A novel self-learning fault detection system for gas turbine engines
Author :
Patel, V.C. ; Kadirkamanathan, V. ; Thompson, H.A.
Author_Institution :
Sheffield Univ., UK
Abstract :
Complex machinery such as aircraft gas turbine engines require complex control systems and control algorithms to allow them to operate efficiently and safely over a wide range of operating conditions. This increased complexity makes the task of finding faults extremely difficult. Thus, on occasions present fault detection systems indicate faults in components which cannot be found on later inspection. These occurrences are known as no fault found conditions and result in loss of revenue and profits through unnecessary maintenance actions and delays. This paper focuses on adapting the growing qualities of resource allocating networks to develop a self-learning fault detection system. It is shown that the proposed system is capable of learning new faults and improving its generalising qualities by adapting itself when presented with similar faults to those previously encountered.
Keywords :
aerospace computing; aircraft; fault diagnosis; feedforward neural nets; gas turbines; learning systems; mechanical engineering computing; aircraft; gas turbine engines; learning; neural networks; radial basis function network; resource allocating networks; self-learning fault detection system;
Conference_Titel :
Control '96, UKACC International Conference on (Conf. Publ. No. 427)
Print_ISBN :
0-85296-668-7
DOI :
10.1049/cp:19960666